PanglaoDB: a web server for exploration of mouse and human single-cell RNA sequencing data.
Franzén O, Gan LM, Björkegren JLM. · 2019
Read PanglaoDB: a web server for exploration of mouse and human single-cell RNA sequencing data. by Franzén O, Gan LM, Björkegren JLM. (2019) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc6450036.txt.
Introduction
Franzén and co-authors introduce PanglaoDB, a free web database that makes published mouse and human single-cell RNA sequencing data explorable without heavy computation. Hundreds of datasets existed as raw files that biologists could not easily open. PanglaoDB pre-processes them into a uniform, browsable form.
The resource covers more than 1,054 experiments and over 4 million cells across major platforms and tissues. Visitors can query cell types, pathways, and regulatory networks directly in the browser. A community-curated marker collection with more than 6,000 gene-to-cell-type associations supports automatic labeling of clusters. Atlas builders gain a shared reference instead of reprocessing the literature study by study.
Catalog record
| CREATOR | Franzén O, Gan LM, Björkegren JLM. |
|---|---|
| YEAR | 2019 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 37 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
PanglaoDB: a web server for exploration of mouse and human single-cell RNA sequencing data By Oscar Franzén, Li-Ming Gan, Johan L M Björkegren Published in: Database: The Journal of Biological Databases and Curation DOI: 10.1093/database/baz046 ABSTRACT Abstract Single-cell RNA sequencing is an increasingly used method to measure gene expression at the single cell level and build cell-type atlases of tissues. Hundreds of single-cell sequencing datasets have already been published. However, studies are frequently deposited as raw data, a format difficult to access for biological researchers due to the need for data processing using complex computational pipelines. We have implemented an online database, PanglaoDB, accessible through a user-friendly interface that can be used to explore published mouse and human single cell RNA sequencing studies. PanglaoDB contains pre-processed and pre-computed analyses from more than 1054 single-cell experiments covering most major single cell platforms and protocols, based on more than 4 million cells from a wide range of tissues and organs. The online interface allows users to query and explore cell types, genetic pathways and regulatory networks. In addition, we have established a community-curated cell-type marker compendium, containing more than 6000 gene-cell-type associations, as a resource for automatic annotation of cell types. INTRODUCTION Single-cell RNA sequencing (scRNA-seq) is a technology that measures gene expression at the single-cell level (1). scRNA-seq achieves an unprecedented level of resolution and it is increasingly used to examine the cellular architecture of tissues, organs and whole organisms. In contrast to bulk RNA-seq, where gene expression is measured and averaged across thousands of cells, scRNA-seq provides much more detailed information and has generated new insights into cellular states, trajectories and heterogeneities. In a typical scRNA-seq experiment, cells from tissue biopsies are dissociated, RNA is converted to cDNA and libraries are generated containing thousands of transcriptomes. Each transcriptome is tagged using a unique oligonucleotide barcode. Certain sequencing protocols incorporate Unique Molecular Identifiers (UMIs) (2) in their workflows so that PCR duplicates can be removed at the data analysis stage. Compared with bulk RNA-seq, scRNA-seq data contain many zero measurements, caused by dropout events and a cleaner biological signal (3). Several protocols and platforms have been developed for scRNA-seq, for example Drop-seq (4), 10X Chromium and SMART-seq2 (5). The rapid rise of scRNA-seq has led to the accumulation of massive amounts of sequencing data in public archives [such as the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA)], since most journals and funders require that upon publication, sequencing data are released to the public domain. However, deposited data often remain difficult to access as it requires significant pre-processing to become useful for regular analysis. Moreover, while the NCBI SRA is an excellent resource for data storage, there is little to no mechanism for quality control, data curation and annotation. Quick access to published datasets allows researchers to answer new questions using old data, prevents duplication of previous efforts and perhaps most importantly, enables comparisons with in-house data to validate or generate new biological hypotheses. Altogether, there is a strong need from the scientific community for efforts involving collection, curation and integration of scRNA-seq data with bioinformatic workflows into platforms that are easily accessible. Previous efforts to develop integrative databases for scRNA-seq analysis include scRNASeqDB (6) and SCPortalen (7), the former being limited to 36 pre-processed datasets collected from the Gene Expression Omnibus (GEO) (8). SCPortalen appears relatively limited in scope and does not provide advanced visualization tools since it is more focused on the technical properties of scRNA-seq data. None of the databases provide pre-computed bioinformatic analyses and advanced visualization from a user perspective. Here, we have developed PanglaoDB—a protocol-agnostic platform for the exploration of scRNA-seq data through a web-based interface. We have collected data and metadata from hundreds of human and mouse studies and processed these data through a unified computational pipeline. In addition to enabling exploration of scRNA-seq experiments, our database provides a manually curated list of cell-type markers that can be incorporated into novel algorithms for inference of cell types. The aim of our work is to provide a frequently updated online single-cell resource to facilitate investigation and hypothesis-free exploration of scRNA-seq data generated by independent academic labs around the world. PanglaoDB unlocks access to more than 1000 single cell experiments, and as such represents the most up to date public resource of curated scRNA-seq data ready for open use by the scientific community. MATERIALS AND METHODS Web server and interface The database is hosted on a Virtual Private Server running Ubuntu Linux with four virtual CPUs, 16 GB RAM and 500 GB hard drive space. We decided to use Nginx as web server because it is relatively lightweight and memory-lean. Nginx was configured to use an SSL certificate from Let’s Encrypt. MySQL was used to keep track of data processing steps and leverage the database through the web interface. The interactive view was built using the D3.js JavaScript library and Python scripts for pulling data. Data collection and bioinformatics pipeline Experimental metadata from high-throughput sequencing studies were downloaded from the NCBI SRA (9) (ftp-trace.ncbi.nlm.nih.gov/sra/reports/Metadata/). We used only submissions fulfilling the following criteria: (i) listed without controlled access; (ii) classified as transcriptomic; and (iii) species is human (TaxID = 9606) or mouse (TaxID = 10090). We then searched abstracts, titles and sample identifiers using the following, case insensitive, regular expression: /(single cell seq|drop\-*seq|scrna|single cell rna-seq|10x\s*(genomics|chromium)|smart-seq2)/. The sequence data were then examined to determine if barcodes and/or UMIs were encoded in the submission; submissions without proper barcode information were discarded from further processing. Submissions that passed the filtering criteria were manually inspected to make sure that these were true scRNA-seq studies. In SRA terminology, each submission receives a unique identifier (/SRA[0-9]+/). A submission may consist of more than one sample (/SRS[0-9]+/), which may consist of more than one sequencing run (/SRR[0-9]+/). In PanglaoDB, a sample is a dataset consisting of gene expression measurements from cells originating from a common biological source or experiment. Multiple runs for each sample were merged into one file. Some studies prefer to deposit each cell with their own SRS accession, in those cases the sample is referred to by their SRA accession. We used the prefetch program in the sratoolkit v.2.9.0 to streamline downloading of the sequencing data. Submissions to the ERA and DRA databanks were not included due to incompatibility with prefetch. Sequencing reads were extracted from SRA files using either fastq-dump or vdb-dump. For data generated using 10X Chromium, barcodes and UMIs were assumed to be 16 and 10 bp in length, respectively. For data generated using Drop-seq, barcodes and UMIs were assumed to be 12 and 8 bp in length, respectively. Barcode and UMI sequences were appended to the read identifier, which is the first line in a FASTQ record: @[readid]_[barcode]_[UMI]. SAM to BAM conversion, sorting and indexing were performed with samtools (10) v.1.8 or sambamba (11) v.0.6.7. The featureCounts program of subread (12) v.1.6.2 was used to add gene information to the BAM file. If the single-cell protocol used UMIs, we performed UMI deduplication using UMI-tools (13) v.0.5.3 (parameters: “--no-sort-output --method unique --gene-tag=XT --per-gene –per-cell”). All sequencing reads were mapped with hisat (14) v.2.1.0, which was selected because of its low memory footprint; only reads with mapping quality